FAMILY_FERTILIZER_WEATHER_STRESS_SYNERGIES: Experiment Log

Overview

Testing compound multiplicative effects between input cost shocks (NPK fertilizer, diesel, electricity) and weather stress conditions (drought, heat) on Dutch potato prices. This hypothesis uses REAL DATA ONLY from Eurostat APIs, Open-Meteo weather, and BoerderijApi prices to measure perfect storm dynamics where stressed crops require more inputs while input costs are elevated.

Hypothesis Origins

• FAMILY_WEATHER_ACCUMULATION: 92.4-97.1% improvement through cumulative weather stress - provides weather stress foundation
• FAMILY_INPUT_COST_TRANSMISSION: Established 4-8 week input cost transmission lags but missed weather synergies
• 2022 Perfect Storm: NPK crisis (300% spike) + severe drought created unprecedented price volatility - natural experiment
• 2024 Storage Crisis: Energy costs + heat stress forced storage releases - compound pressure validated
• Industry Evidence: Farmers apply extra fertilizer to drought-stressed crops at peak cost periods
• Academic Basis: Plant Physiology (2024) compound stress multiplicative effects 65-80% vs 20-35% individual

Experiment Design

• Method: Rolling-origin cross-validation with compound stress period analysis
• Training Window: 365 days minimum for seasonal patterns
• Step Size: 7 days (weekly alignment with price data)
• Test Window: 60 days maximum for both horizons
• Baselines: ALL 4 mandatory standard baselines (persistent, seasonal_naive, ar2, historical_mean)
• REAL DATA ONLY: Eurostat APRI_PI15_INQ + NRG_PC_204/205 + Open-Meteo + BoerderijApi

Data Sources (REAL DATA ONLY - NO SYNTHETIC DATA ALLOWED)

• NPK Fertilizer: Eurostat APRI_PI15_INQ quarterly fertilizer price indices - git:current
• Diesel Prices: Eurostat NRG_PC_204 monthly diesel prices (transport costs) - git:current
• Electricity: Eurostat NRG_PC_205 industrial electricity prices (storage cooling) - git:current
• Weather Stress: Open-Meteo API temperature, precipitation, soil moisture - git:current
• Potato Prices: BoerderijApi NL.157.2086 consumption potatoes weekly - git:current

Compound Stress Definitions

• NPK Shock: >15% increase over 6 weeks OR >20% over 8 weeks (quarterly data alignment)
• Energy Shock: >10% diesel OR electricity increase over 2-4 weeks
• Drought Stress: SPI-3 < -1.0 AND soil moisture < 25th percentile for 2+ weeks
• Heat Stress: Temperature >25°C for 3+ consecutive days OR >30°C single day
• Perfect Storm: All three stress conditions (NPK + energy + weather) TRUE simultaneously

Experiment Runs

Variant A: NPK Shock During Drought Stress (Multiplicative Effects)

Status: Pending - Model: GradientBoosting with drought-fertilizer multiplicative interactions - Key Feature: npk_shock_drought_interaction = (1 + drought_stress) × (1 + npk_shock) - Horizons: 30-day, 60-day - Innovation: Captures elevated fertilizer demand during crop stress periods - Expected: 10-15% improvement over strongest baseline - CRITICAL: Must use REAL Eurostat APRI_PI15_INQ NPK data + Open-Meteo drought indices

Variant B: Energy Cost Spike During Heat Stress (Compound Storage Pressure)

Status: Pending
- Model: RandomForest with heat-energy compound pressure features - Key Features: storage_cooling_costs, transport_stress_costs, compound_energy_stress - Horizons: 30-day, 60-day - Innovation: Energy costs spike exactly when cooling/transport needs are highest - Expected: 12-18% improvement over strongest baseline - CRITICAL: Must use REAL Eurostat NRG_PC_204/205 energy data + Open-Meteo heat stress

Variant C: Triple Stress Interactions (Perfect Storm Conditions)

Status: Pending - Model: XGBoost with triple multiplicative interaction terms - Key Feature: triple_stress_multiplier = (1 + npk_stress) × (1 + energy_stress) × (1 + weather_stress) - 2 - Horizons: 30-day, 60-day - Innovation: Perfect storm amplification effects beyond sum of individual stressors - Expected: 15-22% improvement over strongest baseline - CRITICAL: ALL inputs from REAL data sources - NPK + energy + weather simultaneous stress

Statistical Tests

• Diebold-Mariano test with Harvey-Leybourne-Newbold correction vs ALL 4 standard baselines
• TOST equivalence test with SESOI = 10% improvement threshold
• Markov-switching regime test for compound stress vs normal periods
• Multiplicative effects test: compound model vs additive baseline
• FDR correction for multiple comparisons across variants
• MANDATORY: Compare against strongest baseline (lowest error among 4 standard baselines)

Perfect Storm Analysis

• Temporal coincidence detection: Input shocks + weather stress overlap periods
• Multiplicative amplification: Compound effects vs sum of individual effects
• Stress timing analysis: Production cycle phase impacts (tuber initiation/bulking)
• Market panic proxy: Multiple simultaneous stress signals
• Historical validation: 2022 fertilizer crisis + drought natural experiment

Success Criteria

• Statistical Significance: DM test p<0.05 vs strongest baseline
• Practical Significance: >10% improvement (SESOI threshold)
• Directional Accuracy: >65% correct price direction predictions
• Compound Amplification: Multiplicative > additive effects during stress periods
• Perfect Storm Detection: Higher improvements during triple stress periods

Validation Periods

Natural Experiments (Historical Perfect Storms)

• 2022 Crisis: NPK prices +300%, severe drought, record potato prices €45/100kg
• 2024 Storage Crisis: Energy costs peak + heat stress + early storage releases
• 2018 Drought: Extreme heat + moderate input cost pressures

Control Periods (Normal Conditions)

• 2019: Stable input costs, normal weather patterns
• 2021: Moderate conditions for baseline comparison
• 2020: Single stress factors (no compound effects)

Risk Factors

• Quarterly NPK Data: Eurostat APRI_PI15_INQ quarterly frequency may limit temporal resolution
• Stress Coincidence: Perfect storm events may be rare, requiring longer time series
• Regime Stability: Compound stress relationships may be non-stationary
• Data Alignment: Quarterly fertilizer data with weekly price/weather data synchronization

Verdicts

Variant A: NPK Shock During Drought Stress

Status: Pending - Expected completion after real data implementation - No synthetic results to be included

Variant B: Energy Cost Spike During Heat Stress

Status: Pending - Expected completion after real data implementation - No synthetic results to be included

Variant C: Triple Stress Interactions

Status: Pending
- Expected completion after real data implementation - No synthetic results to be included

HE Notes

• Created 2025-08-19 building on FAMILY_WEATHER_ACCUMULATION success (92-97%) and FAMILY_INPUT_COST_TRANSMISSION foundations
• First hypothesis to systematically model compound multiplicative stress effects in potato markets
• 2022 fertilizer crisis + drought provides natural experiment validation opportunity
• All variants use ONLY REAL DATA from verified repository interfaces
• SESOI set to 10% reflecting strong compound mechanism theoretical foundation
• Perfect storm detection capability for extreme market volatility periods
• Critical innovation: multiplicative rather than additive stress modeling

Decision Log

2025-08-19: Initial Formulation (HE)

• Innovation: First systematic compound stress hypothesis in repository
• Foundation: Builds on proven FAMILY_WEATHER_ACCUMULATION (97% success) + input cost mechanisms
• Data Sources: All REAL data confirmed available through repository interfaces
• Key Insight: Perfect storm conditions create multiplicative rather than additive effects
• Next Steps: EX implementation required with mandatory 4 standard baselines
• Risk Mitigation: Quarterly fertilizer data offset by strong weather/energy monthly data

(Experiment results to be updated after implementation with REAL DATA ONLY)